AlArnaout Zakwan, Zaki Chamseddine, Kotb Yehia, AlAkkoumi Mouhammad, Mostafa Nour
College of Engineering and Technology, American University of the Middle East, 54200, Egaila, Kuwait.
Sci Rep. 2025 Jul 10;15(1):24898. doi: 10.1038/s41598-025-08582-2.
Driver drowsiness is a critical issue in transportation systems and a leading cause of traffic accidents. Common factors contributing to accidents include intoxicated driving, fatigue, and sleep deprivation. Drowsiness significantly impairs a driver's response time, awareness, and judgment. Implementing systems capable of detecting and alerting drivers to drowsiness is therefore essential for accident prevention. This paper examines the feasibility of using heart rate variability (HRV) analysis to assess driver drowsiness. It explores the physiological basis of HRV and its correlation with drowsiness. We propose a system model that integrates wearable devices equipped with photoplethysmography (PPG) sensors, transmitting data to a smartphone and then to a cloud server. Two novel algorithms are developed to segment and label features periodically, predicting drowsiness levels based on HRV derived from PPG signals. The proposed approach is evaluated using real-driving data and supervised machine learning techniques. Six classification algorithms are applied to labeled datasets, with performance metrics such as accuracy, precision, recall, F1-score, and runtime assessed to determine the most effective algorithm for timely drowsiness detection and driver alerting. Our results demonstrate that the Random Forest (RF) classifier achieves the highest testing accuracy (86.05%), precision (87.16%), recall (93.61%), and F1-score (89.02%) with the smallest mean change between training and testing datasets (-4.30%), highlighting its robustness for real-world deployment. The Support Vector Machine with Radial Basis Function (SVM-RBF) also shows strong generalization performance, with a testing F1-score of 87.15% and the smallest mean change of -3.97%. These findings suggest that HRV-based drowsiness detection systems can be effectively integrated into Advanced Driver Assistance Systems (ADAS) to enhance driver safety by providing timely alerts, thereby reducing the risk of accidents caused by drowsiness.
驾驶员疲劳驾驶是交通系统中的一个关键问题，也是交通事故的主要原因。导致事故的常见因素包括酒后驾车、疲劳和睡眠不足。疲劳会严重损害驾驶员的反应时间、意识和判断力。因此，实施能够检测并提醒驾驶员疲劳的系统对于预防事故至关重要。本文研究了使用心率变异性（HRV）分析来评估驾驶员疲劳的可行性。探讨了HRV的生理基础及其与疲劳的相关性。我们提出了一个系统模型，该模型集成了配备光电容积脉搏波描记法（PPG）传感器的可穿戴设备，将数据传输到智能手机，然后再传输到云服务器。开发了两种新颖的算法来定期分割和标记特征，根据从PPG信号中得出的HRV预测疲劳程度。使用实际驾驶数据和监督机器学习技术对所提出的方法进行评估。将六种分类算法应用于标记的数据集，评估诸如准确率、精确率、召回率、F1分数和运行时间等性能指标，以确定用于及时检测疲劳和提醒驾驶员的最有效算法。我们的结果表明，随机森林（RF）分类器实现了最高的测试准确率（86.05%）、精确率（87.16%）、召回率（93.61%）和F1分数（89.02%），训练和测试数据集之间的平均变化最小（-4.30%），突出了其在实际应用中的稳健性。具有径向基函数的支持向量机（SVM-RBF）也显示出很强的泛化性能，测试F1分数为87.15%，平均变化最小，为-3.97%。这些发现表明，基于HRV的疲劳检测系统可以有效地集成到高级驾驶员辅助系统（ADAS）中，通过及时发出警报来提高驾驶员安全性，从而降低因疲劳导致的事故风险。
